socioeconomic inequalities in cancer - stata

Socioeconomic inequalities in cancer: using population‐based cancer registry data

and vital statistics

がんにおける社会経済指標による格差:がん登録、⼈⼝動態統計を⽤いた解析Yuri ItoSenior Researcher

伊藤ゆり主任研究員

2017 Japanese Stata Users Group15:00‐15:50, 16th Sep. 2017

@Kyoto Research Park, Kyoto, Japan

Cancer Control CenterOsaka International Cancer Institute

⼤阪国際がんセンターがん対策センター

©Yuri Ito, CCC OICI

Health inequalities• Inequalities in health is global

concern.• Most developed countries

monitor the health inequalities using official statistics.


In Japan…

The target of the second edition of Healthy Japan 21 was“Reduce the health inequalities”


Socioeconomic status (SES) Health outcome

e.g. Income, Occupation, Education e.g. mortality, incidence, survival

Ideally…Individual record linkage between SES and Health outcome is needed.

However…In Japan, individual linkage between SES and Health outcome using official statistics has not been implemented yet.

To evaluate the causal relationship of those, cohort study is desirable.For health policy, we need to monitor long‐term trends of socioeconomic inequalities by using routinely collected data, covering all population.


Background• Long‐term economic recession in Japan

→ Growing social inequalities→ Inequalities appeared in health

• The target of the second edition of Healthy Japan 21 was“Reduce the health inequalities”

• The 3rd Osaka Cancer Plan is following the target• But the system to monitor the health inequalities has not

been established yet

Objectives• Monitor the socioeconomic inequalities in cancer using

population‐based data Japan


Prevention

Prevention

Screen-ing

Screen-ing

Diag-nosisDiag-nosis

Treat-mentTreat-ment

Palliative care

Palliative care

CureCure

がんにならない

たばこ対策肝炎対策

早く⾒つける

がん検診胃・⼤腸・肺乳房・⼦宮頸

確実に診断

要精検者を精密検査に

がん医療の均てん化

Cancer Control ContinuumTrends in incidence and mortality

Cancer Survival & Cure fraction

% of early stage diagnosis

Cancer registry data is important to evaluate cancer control activities

（就労⽀援）

看取りの場所


Cancer Outcome• Mortality

o Total impact on the society

• Incidenceo Risk factoro Screening

• Survivalo Cancer Care

Population‐based Cancer Registryがん登録が必要！

Vital statistics , ⼈⼝動態統計


Inequalities in cancer• Gaps between…

o Gendero Ageo Area

• Prefecture, second medical areao Socio‐economic status (SES)

• Areal deprivation gap


Areal deprivation index • The small areas based on patients’ address

linkage• Estimated by areal aggregated census‐based information:o Family type: Proportion of elderly couples, elderly singles and mother‐

child familyo Housing type: Proportion of rented accommodationo Occupation type: Proportion of Clerical, Sales, Agriculture,

Production/Labour, and Unemployed

• Areal deprivation index (0‐100) were divided into quintiles (Q1: least deprived, …, Q5: most deprived)

Nakaya et al. PLoS One. 2014

xtile dep5x = dep [fw=pop_all], nq(5)


JGSS (Sample survey)

National Census

Areal‐level deprivation index were estimated by using same items in two database

Areal Deprivation Index：ADI

Incidence and survival: address of patients from cancer registry data (Cho‐Aza level, small)

Mortality: Vital statistics (municipality level, large)

Nakaya et al. PLoS One. 2014

merge 1:1 citycode


Socioeconomic inequalities in Cancer Survival in Osaka, Japan

Ito Y, Nakaya T, Nakayama T, et al. Acta Oncol. 2014;53:1423‐33.


Data sources• Osaka Cancer registry• 237,848 patients who were diagnosed as major 13 sited of cancer in 1993‐2004

Esophagus, Stomach, Colorectum, Liver, Gallbladder, Pancreas, Larynx, Lung, Breast, Cervix uteri, Corpus uteri, Ovary, Prostate, Bladder

• Linked to the ADI based on the patients address (Cho‐Aza level)


Methods• Net survival: 1‐year, 5‐year and Conditional 5‐year

o Pohar‐Perme estimates o By sex, cancer site, period at diagnosis, quintile of ADI

• Variance‐weighted least square regression modelModel 1

Model 2

NS: net survival 1,2,3 period at diagnosis ,

1,2,3,4,5 Quintile deprivation group

stns (SJ14-1: st0326)

vwlsest store& estout


020

4060

8010

0相

対⽣

存率

(%

)

0 1 2 3 4 5診断からの年数

Net survival curves by quintile of ADIStomach cancer, male, 1993‐1996

1 Least deprived

5 Most deprived

234

ADI quintileSocio‐EconomicStatus (SES)

High

Low

Years since diagnosis

Net su

rvival (%

)


Deprivation gap in 1‐year and 5‐year net survival in Osaka, Japan: 1993‐2004

Bladder

Prostate

Lung

Larynx

Pancreas

Liver

Colorectum

Stomach

Esophagus

-20 -15 -10 -5 0 5%

1-year net survival (Men)

BladderOvary

CorpusCervixBreast

LungPancreas

LiverColorectum

StomachEsophagus

-20 -15 -10 -5 0 5%

1-year net survival (Women)

Bladder

Prostate

Lung

Larynx

Pancreas

Liver

Colorectum

Stomach

Esophagus

-20 -15 -10 -5 0 5%

5-year net survival (Men)

BladderOvary

CorpusCervixBreast

LungPancreas

LiverColorectum

StomachEsophagus

-20 -15 -10 -5 0 5%

5-year net survival (Women)

twoway bar & rcap


Pancreas

Lung

Esophagus

Liver

Stomach

ColorectumBladder

Larynx

Prostate

Pancreas

Lung

Liver

Esophagus

Stomach

Ovary

BladderColorectum

Cervix

UterusBreast

-20

-15

-10

-5

0

5

0 50 100 0 50 100

Men WomenAb

solu

te d

epriv

atio

n ga

p (%

)General life tables

One-year net survival

Figure 1a. Deprivation gap in one-year net survival

twoway scatter & rcaplfit


Pancreas

Lung LiverEsophagus

Stomach Colorectum

Bladder

Larynx Prostate

PancreasLiver

LungEsophagus

Ovary

Stomach

Bladder

ColorectumCervix

Uterus

Breast

-20

-15

-10

-5

0

5

0 50 100 0 50 100

Men WomenAb

solu

te d

epriv

atio

n ga

p (%


Five-year net survival

Figure 1b. Deprivation gap in five-year net survival


Pancreas

Lung

LiverEsophagus

Stomach

ColorectumLarynx

Bladder

Prostate

Pancreas

Liver

Lung

Esophagus

Ovary

Stomach

ColorectumCervix

Bladder

Uterus

Breast

-20

-15

-10

-5

0

5

0 50 100 0 50 100

Men WomenAb

solu

te d

epriv

atio

n ga

p (%


Conditional five-year net survival

Figure 1c. Deprivation gap in conditional five-year net survival


Change of deprivation gap in period at diagnosis

• Reduced gapo Pancreatic cancer in men: 1‐year net survivalo Stomach cancer in women: 1‐year net survival

• Widened gapo Lung cancer in men: 1‐ and 5‐year net survival

• No change in other cancer sites


Summary: deprivation gap in cancer survival• Wider gaps were observed in cancer sites which can be

earlier detectedo Differences in % of screened

Tabuchi et al. Nihon‐Iji‐Shinpo (in Japanese). (4605), 84‐88,2012‐7‐28

o Deprived people diagnosed at later stage• Not so many changes in deprivation gap

‐> Need to monitor after 2005 (started widening economic gap)

• Further studies for differences in cancer care access and the treatment (cost) were needed‐> Need to record linkage


Socio‐economic differences in stage‐specific cancer incidence in Osaka, Japan: 1993‐2004

Ito Y, Nakaya T, Kondo N et al (in preparation)


Why Stage‐specific incidence?• Variation in cancer incidence related to prevalence of

cancer risk factor and early diagnosis• Early diagnosis of cancer was influenced by

o Better access to screening programmesor

o Awareness of cancer symptoms• When we monitor socio‐economic differences in cancer

incidence, we need to monitor early‐ and late‐stage incidence separately to consider the influence of early detection

SES


Methods• 164,402 cases: stomach, colorectal, lung, breast, cervical and prostate cancer

• Osaka Cancer Registry, Japan (Pop. 8.8 mil.)

• Age‐standardised incidence rate (Std.Pop. 1985)o Period at diagnosis: 1993‐1998/1999‐2004o Stage at diagnosis (SEER summary stage 2000)

• Early stage: Localised• Late stage: Regional and Distant metastasis

o Socio‐economic status: Areal deprivation index

dstdize


05

1015

2025

30

05

1015

2025

30

Leas

t dep

rived 2 3 4

Most d

epriv

edLe

ast d

epriv

ed 2 3 4Mos

t dep

rived

1993-1998 1999-2004Ag

e-st

anda

rdis

ed in

cide

nce

rate

/100

,000

Regional/Distant

Deprivation gap1. We applied variance weighted least square

regression

2. Deprivation gap= Absolute difference between Q5 – Q1

3. Examined change in gap using interaction between deprivation and period

vwls


05

1015

05

1015

Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5

1993-1998 1999-2004

AS

IR/1

00,0

00

Early stage

05

1015

05

1015

Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5

1993-1998 1999-2004

AS

IR/1

00,0

00

Late stage

Stomach, women

05

1015

2025

30

05

1015

2025

30

Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5

1993-1998 1999-2004

AS

IR/1

00,0

00

Early stage

05

1015

2025

30

05

1015

2025

30

Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5

1993-1998 1999-2004

AS

IR/1

00,0

00

Late stage

Stomach

Men

Women

twoway bar & line


05

1015

05

1015

Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5

1993-1998 1999-2004

AS

IR/1

00,0

00

Early stage

05

1015

05

1015

Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5

1993-1998 1999-2004

AS

IR/1

00,0

00

Late stage

Stomach, women

Lung

Men

Women

twoway bar & line


前⽴腺がん

05

1015

05

1015

Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5

1993-1998 1999-2004

AS

IR/1

00,0

00

Early stage

05

1015

05

1015

Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5

1993-1998 1999-2004

AS

IR/1

00,0

00

Late stage

05

1015

05

1015

Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5

1993-1998 1999-2004

AS

IR/1

00,0

00

Early stage (incl. in situ)

05

1015

05

1015

Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5

1993-1998 1999-2004

AS

IR/1

00,0

00

Late stage

Cervix

Prostate


Prostate

Lung

CRC (insitu)

Stomach

-6.0 -4.0 -2.0 0.0 2.0

1993-1998 1999-2004

Early stage

Prostate

Lung

CRC

Stomach

-5.0 0.0 5.0 10.0

1993-1998 1999-2004

Late stage

*

*

* *

*

*

Summary of deprivation gap: Men

*p<0.05 for interaction of period and deprivation index

twoway bar & rcap


Cervix (insitu)

Breast (insitu)

Lung

CRC (insitu)

Stomach

-2.0 -1.0 0.0 1.0 2.0 3.0

1993-1998 1999-2004

Early stage

Cervix

Breast

Lung

CRC

Stomach

0.0 1.0 2.0 3.0 4.0 5.0

1993-1998 1999-2004

Late stage

*

Summary of deprivation gap: Women

*p<0.05 for interaction of period and deprivation index

twoway bar & rcap


Summary of results• Late stage incidence

o More deprived patients > less deprived patients, except for prostate cancer

o Gaps became narrower• Early stage incidence

o Less deprived patients > more deprived patients, especially for male stomach, colorectal, prostate cancer

o Gaps became wider during recent period• Cervical cancer for both stages

o More deprived patients > less deprived patients


Possible explanation• Inverse contrast of early stage incidence in men

o Difference in the screening type related to working place (Health insurance is determined the style of cancer screening in Japan)

• High‐risk health behaviour in deprived groupo Tobbaco smoking, risky sexual activitieso Less participation to the screeningo Lack of awareness and delay in access to cancer care


Socioeconomic Inequalities in Cancer Mortality in Japan

政策科学総合研究事業（統計情報総合研究）「健康格差対策に必要な公的統計のあり⽅に関する研究」班


Data Sources• Vital statistics data for the period between 2005 and 2014, including residential area (municipality‐level)

• Population data by sex, 5‐year age group and municipality: National Census for 2006 and 2010‐> Population data for other years wereinterpolated/extrapolated

• The areal deprivation index (ADI):o Socio‐economic status based on census‐based information by municipality (Nakaya T. et al. PLoSOne. 2014)

merge 1:1 citycode


Trends in all cancers by ADI quintile

Q5: Most deprived

Q1: Least deprived

twoway scatter & line


Statistical Analysis1. Age‐standardised mortality rate (ASMR) by

quintile group of ADI ‐> Absolute differences Q5 and Q1 ASMR‐> Attributable fraction of each cause of death

2. Relative risks (RR) of Q1 (ref) to Q2‐Q5 estimated by Poisson regression model

∶ municipality,k:quintile

glm obs dep2-dep5, family(poisson) link(log) offset(lnexp) eform

dstdize


Socio‐Economic Position (SEP)ADI

∶ municipality: proportion of municipality

SEP

↑/2

egen sumpop=sum(pop)gen csumpop=sum(pop)


Statistical Analysis3. Relative Index of Inequalities (RII) by year

glm obs sepi, family(poisson) link(log) offset(lnexp) eformgen coef_sepi=_b[sepi]gen se_sepi=_se[sepi]gen rr_sepi=exp(coef_sepi)gen rr_cil_sepi=exp(coef_sepi-1.96*se_sepi)gen rr_ciu_sepi=exp(coef_sepi+1.96*se_sepi)


Statistical Analysis4. Trends in RII using interaction term

glm obs sepi i.cnyear i.cnyear##c.sepiif sex==`g', family(poisson) link(log) offset(lnexp) eform


.91

1.1

1.2

.91

1.1

1.2

1 2 3 4 5 1 2 3 4 5

Male FemaleStan

dard

ised

Mortality Ra

tio2006‐20092010‐2014

Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5

Standardised Mortality Ratios (SMRs) of All sites of cancer death by ADI in Japan in 2006‐2014

* p<0.05 for interaction between ADI and period

*


.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

Oesophagus Stomach Colorectum Liver Pancreas

Larynx Lung Prostate Lymphoma Leukaemia

Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5


* * **

*

* *

Standardised Mortality Ratios （SMRs） of ADIgroup (ref: Q1) in Japan in 2006‐2014: Male

twoway bar & rcap


Standardised Mortality Ratios （SMRs） of ADIgroup (ref: Q1) in Japan in 2006‐2014: Female

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

.81

1.2

1.4

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5

Oesophagus Stomach Colorectum Liver Pancreas

Lung Breast Cervix uteri Corpus uteri Ovary

Lymphoma Leukaemia



Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5

*

*

*

*

*

twoway bar & rcap


Discussion• Wider gaps in cancer mortality between the most and the

least deprived in men were observed than those in women• Deprivation gaps were wide in Liver, Lung, Cervical cancers

and Leukaemiao Liver: Less access to treatment for Hepatitis virus Co Lung: Higher smoking rate in the more deprived groupo Cervix: Lower attendance rate of cancer screeningo Leukaemia: Regional effect? (e.g. High incidence in the west part of Japan), Expensive and long‐term treatment? (e.g. Born‐marrow transplant)

• Inverse contrast was observed in Breast, Corpus uteri and Ovarian cancero Related to the high risk lifestyle in the less deprived group? (reproductive factor, alcohol consumption)


Cancer

Long‐term trends in Relative Index of Inequalities (RII)

Trends in RII by age group twoway scatter & rcap


Limitation of our data availability • Socioeconomic status is based on areal level

o The size of area is large for mortality data‐> under estimate of deprivation gap

o Different area: National Census and Cancer Registry (time consuming to adjust the difference)‐> use exact address to translate a geocode

• In near future, we need to link the National Census and Cancer Registry/Vital Statistics via individual ID


Further steps to reduce deprivation gap

• Monitor deprivation gap in cancer outcome and related factors in all prefectures in Japano Survival, stage‐specific incidence, mortality, smoking rate, screening rate

↓• Try to understand the mechanism of the deprivation gap

↓• Take measures to reduce the gap

Causal mediation analysis

Model simulation

Gformula, paramed,…


Thank you for your attention!

itou‐[email protected]

Please keep in touch to discuss with Stata and the useful command.


socioeconomic inequalities in cancer - stata

Documents